A Simpler Proof of Jensen’s Coding Theorem

이 글은 Pu 이론에 관한 것이다. Pu 이론은 limit coding을 위한 forcing이다. 이 강의에서는 Pu 이론이 어떻게 구성되어 있는지, 특정 조건에서 Pu 이론이 어떻게 작동하는지를 설명한다.

강의는 다음과 같이 나누어 져 있다.

1. Section One: Successor Coding
이 섹션에서는 successor coding에 대해 다룬다. successor coding은 countable limit cardinal의 reshaped string을 subset으로 변환하기 위한 forcing이다. Pu 이론에서 사용되는 forcing P
강의를 요약하면 다음과 같다.

* Pu 이론은 limit coding을 위한 forcing이다.
* Pu 이론은 countable limit cardinal의 reshaped string을 subset으로 변환하기 위한 것이다.
* Pu 이론에서는 특정 조건에서 Pu 이론이 어떻게 작동하는지를 설명한다.

강의를 요약하면 다음과 같다.

* Pu 이론은 forcing P* Pu 이론은 predense와 reduction을 만족해야 한다.
* Pu 이론은 restrict와 같은 조건을 만족해야 한다.

A Simpler Proof of Jensen’s Coding Theorem

arXiv:math/9211204v1 [math.LO] 24 Nov 1992A Simpler Proof of Jensen’s Coding TheoremSy D. Friedman*M.I.T.Beller-Jensen-Welch [82] provides a proof of Jensen’s remarkable Coding The-orem, which demonstrates that the universe can be included in L[R] for some realR, via class forcing. The purpose of this article is to present a simpler proof ofJensen’s theorem, obtained by implementing some changes first developed for thetheory of strong coding (Friedman [87]).The basic idea is to first choose A ⊆ORD so that V = L[A] and then gener-ically add sets Gα ⊆α+, α O or an infinite cardinal (O+ denotes ω) so that Gαcodes both Gα+ and A ∩α+.

Also for limit cardinals α, Gα is coded by ⟨G¯α|¯α < α⟩.Thus there are two “building blocks” for the forcing, the successor coding and thelimit coding. We modify the successor coding so as to eliminate Jensen’s use of“generic codes” (this improves an earlier modification of this type, due to Welchand Donder).

And we thin out the limit coding so as to eliminate the technicalproblems causing Jensen’s split into cases according to whether or not O# exists.Theorem. (Jensen) There is a class forcing P such that if G is P-generic overV then V [G] |= ZFC + V = L[R], R ⊆ω.

If V |= GCH then P preserves cardinals.It is not difficult to class-generically extend V to make GCH true. And any“reshaped” subset of ω1 can be coded by a real via a CCC forcing.

(See SectionOne below for a definition of “reshaped”.) So it suffices to prove that V can becoded by a “reshaped” subset of ω1, preserving cardinals, assuming the GCH.

Asa first step, force A ⊆ORD such that for each infinite cardinal α, Lα[A] = Hα =all sets of hereditary cardinality less than α.Section One The Successor Coding Rs.Fix an infinite cardinal α. Sα is defined to be a certain collection of “strings”s : [α, |s|) −→2, α ≤|s| < α+. For s to belong to Sα we require that s is “reshaped”.

*Research supported by NSF Grant # 9205530.1

This means that for η ≤|s|, L[A∩α, s ↾η] |= card(η) ≤α. The reshaping of s allowsus to code s by a subset of α, in the manner which we now describe.For s ∈Sα define structures A0s = Lµ0s[A ∩α, s∗], As = Lµs[A ∩α, s∗] asfollows (where s∗= {µs↾η|s(η) = 1}) : If |s| = α then µ0s = α.

For |s| > α,µ0s = S{µs↾η|η < |s|} and in general µs = least p.r. closed ordinal µ greater thanµ0s such that Lµ[A ∩α, s∗] |= card(s) ≤α.

These ordinals are well-defined due tothe reshaping of s.For s ∈Sα we write α(s) = α. Note that if |s| = α(s) then s = ∅; in thiscase we think of s as “labelled” with the ordinal α(s), so that there are distinctsα ∈Sα, α(sα) = α.For later use we also define structures bAs and A′s for s ∈Sα as follows: let ˆµs =largest p.r.

closed µ such that µ = µ0s or Lµ[A ∩α, s∗] |= |s| is a cardinal greaterthan α. Then bAs = Lˆµs[A ∩α, s∗].

The ordinal µ′s and structure A′s are defined inthe same way, except we replace p.r. closure of µ by the weaker condition ω · µ = µ.For s ∈Sα+ write ¯s < s to mean that π(¯s) = s where π :A −→As is anelementary embedding with some critical point α(¯s) < α+ and where π(α(¯s)) = α+.Then π = π¯ss is unique.

Let ¯s ≤s denote ¯s < s or ¯s = s. We have the followingfacts:(a) {α(¯s)|¯s < s} is CUB in α+. (b) If ¯t is a proper initial segment of ¯s then ¯t < π¯ss(¯t) = t and π¯tt = π¯ss ↾A¯t.

(c) As = S{Rng(π¯ss)|¯s < s}.Now for s ∈Sα+ let bs = {¯s|¯s < s}. We use the strings ¯s ∗i with ¯s < s ↾η,i = 0 or 1, to code s(η).

A condition in the successor coding Rs is a pair (u, ¯u)where:1) u ∈Sα2) ¯u ⊆{bs↾η|s(η) = 0}, card (¯u) ≤α in As.To define extension of conditions, we need a couple of preliminary definitions. Wesay that ¯u restrains ¯s ∗1 if ¯s ∈b for some b ∈¯u and ¯s lies on u if u(α(¯s)) = 1and u(⟨α(¯s), η⟩) = ¯s(η) for η ∈Dom(¯s).

Also let ⟨Zγ|γ < α+⟩be an Lα+-definablepartition of the odd ordinals less than α+ into α+ disjoint pieces of size α+. We usethe Zγ’s to code A∩α+ into Gα.

For u ∈Sα, ueven (δ) = u(2δ), uodd (δ) = u(2δ+1).Extension of conditions for Rs is defined by: (u0, ¯u0) ≤(u1, ¯u1) iffu0 extendsu1; ¯u0 ⊇¯u1; ¯u1 restrains ¯s∗1, ¯s∗1 lies on ueven0−→¯s∗1 lies on ueven1; γ < |u1|, γ /∈A,2

δ ∈Zγ, uodd0(δ) = 1 −→uodd1(δ) = 1. Note that Rs ∈As.Lemma 1.1.

Suppose G is Rs-generic over As and let Gα = S{u|(u, ¯u) ∈G forsome ¯u}. Then G, A ∩α+, s belong to Lµs[Gα].Proof.

We can write (u, ¯u) ∈G iffu ⊆Gα and ¯s ∈b ∈¯u, ¯s∗1 lies on Gevenα−→¯s∗1lies onueven and γ < |u|, γ /∈A, δ ∈Zγ, Goddα(δ) = 1 −→uodd(δ) = 1. SoG ∈Lµs[A ∩α+, Gα, s].

And γ ∈A ∩α+ iffGoddα(δ) = 1 for unboundedly manyδ ∈Zγ, so G, A ∩α+ ∈Lµs[Gα, s]. Finally note that for any η < |s|, ¯s lies on Gevenαfor unboundedly many ¯s < s ↾η by a density argument using the fact that forη < |s|, (u, ¯u) ∈Rs, bs↾η is almost disjoint from {u|u extends some ¯s ∗1 restrainedby ¯u}.

So s(η) = 1 iff¯s ∗1 lies on Gevenαfor unboundedly many ¯s < s ↾η. Thuss ↾η can be recovered by induction on η ≤|s|, inside Lµs[Gα].⊣Lemma 1.2.

R

Otherwise we need only observe that R

Suppose (u0, ¯u0) ∈Rs and ⟨Di|i < α⟩are predense on Rs, ⟨Di|i < α⟩∈As.By induction we define conditions (ui, ¯ui) and elementary submodels Mi of As with(ui, ¯ui) ∈Mi+1, for i ≤α. Choose M0 to contain α as a subset and to contain⟨Di|i < α⟩, s, A ∩α+ as elements.

Having defined (ui, ¯ui) and Mi, choose Mi+1 tocontain Mi as a subset and (ui, ¯ui) as an element. Choose (ui+1, ¯ui+1) to extend(ui, ¯ui), meet Di, guarantee that if s(η) = 1, η ∈Mi then ¯s ∗1 lies on ueveni+1 −uevenifor some ¯s < s ↾η, guarantee that if γ ∈A ∩(Mi ∩α+) then uoddi+1(δ) = 1 for someδ /∈dom uoddi, δ ∈Zγ, and finally choose ¯ui+1 to contain all bs↾η with s(η) = 0, η ∈Mi.

The last requirement can be imposed because the facts that |s| has cardinality≤α+ in As, Hα+ ⊆As imply that any subset of |s| of cardinality ≤α belongs toAs.For λ ≤α limit, Mλ = S{Mi|i < λ} and uλ = S{ui|i < λ}, ¯uλ = S{¯ui|i < λ}.By construction, uλ codes A ∩(Mλ ∩α+) as well as ¯s = s ◦π−1 where π is thetransitive collapse map for Mλ. Thus the sequence of ordinals ⟨Mi ∩α+|i < λ⟩iscofinal in |uλ| and belongs to L[uλ], since the entire sequence ⟨Mi|i < λ⟩can be3

recovered in L[uλ], Mi = transitive collapse (Mi). This shows that uλ is reshaped,so (uλ, ¯uλ) is a condition.

Finally note that (uα, ¯uα) is an extension of (u0, ¯u0)meeting each of the Di’s.⊣Corollary 1.4. R

By Lemma 1.2 it suffices to prove ≤α-distributivity in A0s. This is easilyproved by induction on |s|, using Lemma 1.3 at successor stages.⊣Lemma 1.5.

If D ⊆R

We can extend(u, ¯u) to guarantee that for some ¯t < t, ¯u = {bt↾(η+1)|t(η) = 0, η ∈Rng π¯tt}, D, s ∈Rng(π¯tt) and |u| = α(¯t) + 1, u(α(¯t)) = 0, u ∈A¯t↾α(¯t). Let (u∗, ¯u∗) be the leastextension of (u, ¯u ∩As) ∈Rs meeting D. We claim that (u∗, ¯u∗∪¯u) is an extensionof (u, ¯u), and this will prove the lemma.

Clearly γ < |u|, δ /∈A, δ ∈Zγ, u∗odd(δ) =1 −→uodd(δ) = 1, since (u∗, ¯u∗) extends (u, ¯u ∩As). Suppose r < t ↾η, t(η) = 0where η ∈Rng π¯tt and r ∗1 lies on u∗even.

If η < |s| then r ∗1 lies on ueven, asdesired, since (u∗, ¯u∗) extends (u, ¯u ∩As). If α(r) < α(¯t) then |r| < α(¯t) so againr ∗1 lies on ueven since |u| > α(¯t) > |r ∗1|.

If α(r) = α(¯t) then r ∗1 cannot lie onu∗even, by choice of u. Finally if α(r) > α(¯t) then since η ≥|s| we have α(r) > |u∗|by leastness of (u∗, ¯u∗).

So r ∗1 cannot lie on u∗even.⊣Section Two Limit Coding.We begin with a rough indication of the forcing Pu for coding u ∈Sα, α an un-countable limit cardinal, into a subset of α. Pu ⊆Au consists of P

Then4

code u by: u(ξ) = 1 iffpoddβ+ (bξβ+) = 1 for sufficiently large β ∈Card ∩α. Recall thatthe successor coding Rpβ++ makes use of odd ordinals (in the Zγ’s) so the successorand limit codings do not conflict.

For p, q ∈Pu we write p ≤q iffp(β) ≤q(β) inRpβ+ for each β ∈Card ∩α.To facilitate the proofs of extendibility and distributivity for Pu we thin outthe forcing, in a number of ways. For this purpose we need appropriate forms of □and ⋄, in a relativized form.

Jensen observed that his proofs of these principles forL go through when relativized to reshaped strings. Precisely:Relativized □Let S = S{Sα|α an infinite cardinal}.

There exists ⟨Cs|s ∈S⟩such that Cs ∈As and:(a) If α(s) < |s| then Cs is closed, unbounded in µ0s, ordertype (Cs) ≤α(s).If |s| is a successor ordinal then ordertype (Cs) = ω. (b) ν ∈Lim(Cs) −→for some η < |s|, ν = µ0s↾η and Cs ∩ν = Cs↾η.

(c)Let π :⟨A, C⟩Σ1−→⟨A0s, Cs⟩and write crit(π) = ¯α, A = L¯µ[A, ¯s∗]. Ifπ(¯α) = α(s) then L[A, ¯s∗] ⊨|¯s| is not a cardinal > ¯α and(c1) C ∈Lµ[A, ¯s∗] where µ is the least p.r.

closed ordinal greater than ¯µ s.t.Lµ[A, ¯s∗] ⊨card(|¯s|) ≤¯α. (c2) π extends to π′ : A′Σ1−→A′s where A′ = Lµ′[A, ¯s∗], µ′ = largest ordinaleither equal to ¯µ or s.t.

ω · µ′ = µ′ and Lµ′[A, ¯s∗] ⊨|¯s| is a cardinal greater than ¯α. (c3) If ¯α is a cardinal and π(¯α) = α then A = A0¯s and C = C¯s.Relativized ⋄Let E = all s ∈S such that |s| limit and ordertype (Cs) = ω.There exists ⟨Ds|s ∈E⟩such that Ds ⊆A0s and:(a)D ∈bAs ̸= A0s, D ⊆A0s −→{ξ < |s|s ↾ξ ∈E, Ds↾ξ = D ∩A0s↾ξ} isstationary in bAs.

(b) Ds is uniformly Σ1-definable as an element of A′s. (c) If A′s ⊨α++ exists then Ds = ∅.Now we use these combinatorial structures to impose some further restrictionson membership in Pu −P

First some definitions. For p ∈Pu and β ∈Card ∩α,(p)β denotes p ↾Card ∩[β, α), D ⊆P

extends an element of D). And p reduces D below β if every q ≤p can be furtherextended to r such that r meets D and (q)β = (r)β.Requirement A.

(Predensity Reduction) Suppose p ∈Pu −P

(A2) If |u| is a successor ordinal, D ⊆P

p ∈Pu↾ξ.If p belongs to Pu and ξ < |p| then there exists r s.t. p ≤r and |r| = ξ.Requirement C.(Nonstationary Restraint)Suppose Au ⊨α inaccessible andp ∈Pu.

Then there exists a CUB C ⊆α s.t. C ∈Au and β ∈C −→pβ = ∅.The remaining Requirement D will be introduced at a later point when wediscuss strong extendibility at successor stages.Extendibility and distributivity for Pu are stated as follows.

Let q ≤β p signifythat q ≤p and q ↾β = p ↾β. (P

∆−distributivity for P

As the base case |u| = αis vacuous we assume from now on that |u| > α. The following consequences ofpredensity reduction are needed in the proof.Lemma 2.1.

(Chain Condition for P

We may assume that bAu ̸= A0u. Suppose D ⊆P

Then D∗∈bAu.By (∗∗)u and Lemma 1.2, D∗is β-predense for all β ∈Card ∩α. (Use ≤β+-6

distributivity of (P

Thus bypredensity reduction and restriction, D∗∩A0u↾ξ is predense on P

Then D is predense on Pv.Proof. By restriction, if p ∈Pv −Pu then p extends some q in Pu −P

By thechain condition for P

Assume (∗∗)u and |u| a limit ordinal. Then (∗)u holds.Proof.

We first claim that if p ∈P

The base case β1 = β+0 andthe case of β1 a successor cardinal follow easily, using (∗∗)u. If β1 is singular inA0u then we can choose γ0 < γ1 < · · · approximating β1 in length λ < β1 andconsider ⟨Eδ|δ < λ⟩where Eδ = all q meeting each Dβ, λ ≤β < γδ, |q| leastso that ⟨Dβ|β0 ≤β < β1⟩∈A0u↾|q|.

Then we are done by induction.If β1 isinaccessible in A0u then either β1 = α, in which case the result follows directly fromthe second statement of (∗∗)u, or β1 < α, in which case we can factor P

By induction on α, we can extend q to meet all the Dβ’s.Now write Cu = {µ0u↾ξi|i < λ} and choose a successor cardinal β0 < α tobe at least as large as λ and the β given in the statement of (∗)u, if λ < α. Nowinductively define a subsequence ⟨ηj|j < λ0⟩of ⟨ξi|i < λ⟩and conditions ⟨pj|j < λ0⟩as follows. First suppose λ < α.

Let p denote the condition given in the statementof (∗)u. Set p0 = p, η0 = least ξi s.t.

p ∈P

all γ, β0 ≤γ < α, q meets all γ+-predense D ⊆P

The ordinal λ0 is determined by thecondition that ηλ0 is equal to |u|. If λ = α then the definition is the same, except indefining pj+1 require pj+1 ≤β∪ℵi+1 pj where ηj = ξi and only require pj+1 to meetγ+-predense D as above for γ between β ∪ℵi and α.We must verify that pδ as defined above is indeed a condition for limit δ.

(Thereis no problem at successor stages, using Lemma 2.2 and the first paragraph of thepresent proof.) First we show that for γ ∈Card ∩α, pδγ is reshaped.

We need onlyconsider γ ≥β and in case λ = α we need only consider γ ≥β∪ℵi where ηδ = ξi. Byconstruction if γ ∈M ηδγ= Σ1 Skolem hull of γ ∪{p, α} in ⟨A0u↾ηδ, Cu↾ηδ⟩then pδγ isπ[(P

And |pδγ| is Σ1-definably singularized over TC(M ηδγ ). Write TC(M ηδγ ) as⟨A, C⟩.

By genericity and cofinality-preservation for π[(P

If M ηδγ ∩α = γthen pδγ is again reshaped because of Relativized □(c1), (no genericity argumentrequired). Lastly if γ′ = min(M ηδγ ∩(ORD −γ)) < α then use the first argument,but with π[(P

As pδ ↾γ is definable over TC(M ηδγ ) ∈L[A∩γ, pδγ] this amounts to showing that µpδγ is large enough. By (∗∗)u↾ηδ and Lemma2.1 we know that P

Andthus A′[pδγ] ⊨|pδγ| is a cardinal. But by Relativized □(c1), TC(M ηδγ ) appearsrelative to pδγ before the next p.r.

closed ordinal after the height of A′. So pδ ↾γ ∈Apδγ .

If M ηδγ∩α = γ then no genericity argument is required; we only needRelativized □(c1).Requirements B, C are easily checked, the latter using the fact that in case ofα inaccessible in A0u we required pj+1 ≤β∪ℵi+1 pj(ηj = ξi) and therefore can usediagonal intersection of clubs. To check Requirement (A1) note that if M ηδγ ∩α ̸= γthen either pδγ /∈E or Dpδγ = ∅, since A′pδγ ⊨γ++ exists and we can applyRelativized ⋄(c).

If M ηδγ∩α = γ then pδγ ∈E iffu ↾ηδ ∈E by Relativized □(c3) and if these hold then by Relativized ⋄(b), π′[Dpδγ ] = Du↾ηδ, where π′ comes8

from Relativized □(c2). So all we need to arrange is that our initial condition pbe chosen to meet Du, in case u ∈E, and otherwise choose η0 to be at least ξω, sothat u ↾ηδ /∈E for limit δ.⊣Lemma 2.4.

Assume |u| limit and (∗)v, (∗∗)v for v ⊆u, v ̸= u. Then (∗∗)u holds.Proof.

We may assume that bAu ̸= A0u. We need only make a small change in theconstruction of the proof of Lemma 2.3.

Given predense ⟨Di|i < β⟩on (P β such that Di ∩(P

We can guarantee ⟨Di∩(P

Suppose (∗∗)u holds and |u| is a successor ordinal. Then (∗)u holds.Proof.

We may assume that the given p belongs to Av −A0v where v = u ↾(|u|−1).Write Cu = ⟨ξj|j < ω⟩. Now proceed as in the construction of the proof of Lemma2.3, making successive ≤β-extensions below p (where β is given in the statement of(∗)u), p ≥β p0 ≥β p1 ≥β · · · so that pj+1 meets all γ+-predense D ⊆P

⟨pi|i ∈ω⟩then ˆq meets the requirements for being a condition atall γ ∈Card ∩α+ with the exception of γ in C ∪{α}, C = {γ|Mγ ∩α = γ}, Mγ = Σ1Skolem hull of γ∪{p, α} in ⟨Av, Cu⟩. The reason is that for γ ∈α−C, Tγ = TC(Mγ)belongs to Aˆqγ, since Tγ ⊨|ˆqγ| is a cardinal and ˆqγ is generic over Tγ.To make ˆq into a condition q ∈Pu we must do two things.First extendˆqγ+ for γ ≥β so as to code u(|v|) = 0 or 1.

This is easily done as there are noconflicts between the successor and limit codings. Second for γ ∈C we extendˆqγ to qγ = ˆqγ ∗u(|v|).

The only remaining question is whether the reatraint ˆqγ9

will allow us to do this. But γ ∈C −→ˆqγ = ∅since C is contained in the CUBwitnessing Requirement C for ˆq at α.⊣Lemma 2.6.

Suppose (∗)u and (∗∗)v, v ⊆u ̸= v hold and |u| is a successor. Then(∗∗)u holds.Proof.

We must show that if v = u ↾(|u|−1) and p ∈(Pv)β −(P

Then Σpf = {q ∈Pv| ∀β ∈Dom(f), q(β) meets all predenseD ⊆Rpβ+, D ∈Mβ}.Sublemma 2.7. Σpf is dense below p in Pv.Before proving Sublemma 2.7 we establish the Lemma, assuming it.

Choosea limit ordinal λ = ωλ < µv such that ⟨Di|i < ω⟩, Cv ∈Av ↾λ = Lλ[A ∩α, v∗]and Σ1 cof(Av ↾λ) = ω. Choose a Σ1(Av ↾λ) sequence λ0 < λ1 < · · · cofinal inλ such that ⟨Di|i < ω⟩, Cv, x ∈Av ↾λ0 where x is a parameter defining the λi’s.Set M iγ = least M ≺Σ1 Av ↾λi such that γ ∪{x, ⟨Di|i < ω⟩, α, Cv} ⊆M, for eachγ ∈Card+ ∩α.

Define fi(γ) = M iγ.Choose p = p0 ≥p1 ≥· · · successively so that pi+1 meets Di and Σpifi. Setp∗= g.l.b.

⟨pi|i ∈ω⟩. We show that p∗is a well-defined condition.If |v| > αthen thanks to (∗∗)v it will suffice to show that if D ∈M iγ ∩A0v is predense on(P

(For then, p∗γ codes ageneric over the transitive collapse of M iγ ∩A0v.) If |v| = α then instead of P

Note thatPα is cofinality-preserving, by applying (∗∗) at cardinals < α.Choose j ≥i so that for k > j, pk reduces D no further than pj. Let γ′ beleast so that pj reduces D below γ′.

Then γ′ < α by Predensity Reduction for p.If γ′ ≤γ then of course we are done. If γ′ > γ is a double successor cardinal thenwe reach a contradiction since by definition pj+1 reduces D further.

If γ′ = δ+, δ alimit cardinal then by Predensity Reduction at δ, D is reduced below some δ′ < δ,another contradiction. If γ′ is a limit cardinal then the same argument applies,replacing γ′ by (γ′)+.10

Finally we have:Proof of Sublemma 2.7. It suffices to show the following.Strong Extendibility Suppose g ∈Av, g(β) ∈Hβ++ for all β ∈Card ∩(β0, α)and p ∈Pv.

Then there is q ≤β0 p such that g ↾β ∈Aqβ for all β ∈Card ∩(β0, α].For, Strong Extendibility allows us to extend to a condition q such that for allβ ∈Card ∩α, g ↾β ∈Aqβ, where g(β) = f(β) ∩Hβ++. Then successively extendeach q(β) to meet predense D in f(β).We now break down Strong Extendibility into the ramified form in which itwill be proved.

For any µ such that µ0v ≤µ < µv, k ∈ω −{0} and β ∈Card ∩α letM µ,kβ= Σk Skolem hull of β ∪{α} in A∗v ↾µ = ⟨Lµ[A ∩α, v∗, Cv], A ∩α, v∗, Cv⟩. (Notice that this structure is Σ1 projectible to α without parameter.

)SE(µ, k) Suppose p ∈Pv and β0 ∈Card ∩α. Then there exists q ≤β0 p such thatTC(M µ,kβ) ∈Aqβ for all β ∈Card ∩(β0, α).It suffices to prove SE(µ, k) for all µ, k as above.

We do so by induction onµ and for fixed µ, by induction on k. To verify the base case of this induction wemust impose one last requirement on our conditions.Requirement DSuppose p ∈Pv −P

Similarly if µ is a successor, k = 1 thenuse ⟨Σk(A∗v ↾µ −1)|k ∈ω⟩to approximate Σ1(A∗v ↾µ), using the Σf’s, f definableover A∗v ↾µ −1.Suppose k > 1. By induction we can assume that TC(M µ,k−1β) ∈Apβ for largeenough β.

If C = {β < α|β = α∩M µ,kβ} is unbounded in α then successively extendp ↾β for β ∈C so that TC(M µ,kβ′ ) ∈Aqβ′ for β′ < β. There is no problem at limitssince TC(M µ,kβ), C ∩β ∈Apβ for β ∈C.If α is Σk(A∗v ↾µ)-singular then choose a continuous cofinal Σk(A∗v ↾µ)sequence β0 < β1 < · · · below α of ordertype λ0 = cof(α).

Also choose βi+1large enough so that M µ,k−1βi+1|= βi is defined.This is possible since A∗v ↾µ =S{M µ,k−1β|β < α}. Now define N iβ for i < λ0, β < βi to be the Σk Skolem hull of β11

in M µ,k−1βi. Then ⟨TC(N iβ)|β < βi⟩∈Apβi for i < λ0 since it is easily defined fromM µ,k−1βi∈Apβi .

Successively λ0-extend p ↾βi, producing p = p0 ≥λ0 p1 ≥λ0 · · ·where TC(N iβ) ∈Apiβ for β ∈(λ0, βi). This is possible by induction on α, and sinceTC(N iβ) is easily defined from ⟨TC(N i¯β)|¯β < β⟩for limit β < βi.

(We must alsorequire that pi+1 meets Σpifi where fi(β) = N iβ.) pλ is well-defined for limit λ ≤λ0and Apλ0β contains ⟨TC(N iβ)|i < λ0⟩and hence TC(M µ,kβ) for β > λ0.

Then useinduction to fill in on (0, λ0] so that SE(µ, k) is satisfied.Lastly, there is the intermediate case where α is Σk(A∗v ↾µ)-regular but C ={β < α|β = α ∩M µ,kβ} is bounded in α. Then Σk+1(A∗v ↾µ)-cof (α) = ω andwe apply induction to produce p = p0 ≥p1 ≥· · · so that pi+1 ↾[βi, βi+1] obeysSE(µ, k) where β0 < β1 < · · · is a cofinal ω-sequence of successor cardinals belowα.

Let q = g.l.b. ⟨pi|i ∈ω⟩.This completes the proof of Sublemma 2.7 and hence of (∗∗)u.⊣Section Three Proof of Jensen’s TheoremA condition in P is a function p from an initial segment of Card into V suchthat Dom(p) has a maximum α(p), for any α ∈Dom(p), p(α) = (pα, ¯pα), if α ∈Dom(p) ∩α(p) then p(α) belongs to Rpα+, p(α(p)) = (s(p), ∅) where s(p) ∈Sα(p)and for uncountable limit cardinals α ∈Dom(p), p ↾α ∈Ppα.

And q ≤p in P ifα(p) ≤α(q), s(p) ⊆qα(p) and for α ∈Dom(p) ∩α(p), q(α) ≤p(α) in Rqα+.For any α ∈Card, s ∈Sα, Ps denotes all p ↾α for p ∈P such that α(p) = αand s(p) = s. And Pα denotes all p ∈P such that α(p) < α.Now suppose α is an uncountable limit cardinal and s ∈Sα, |s| = α + 1. ByLemma 2.2, G P-generic −→G ∩P

closed ordinal greater than α. As the forcing relation for P

closed ordinal> α.Now note that P preserves cofinalities, as otherwise Ps would change cofi-nalities for some s as above, contradicting Distributivity (Lemmas 2.4, 2.6) andChain Condition (Lemmas 1.2, 2.1). If G is P-generic then L[G] = L[X] where12

X = Gω ⊆ω1. Finally by Jensen-Solovay [68], X can be coded by a real via aCCC forcing.

This completes the proof of Jensen’s Coding Theorem, subject to theverification of Relativized □and ⋄.Section Four Relatived Square and DiamondFor completeness, we prove Relativized □and ⋄. As relativization causes no se-rious problems, we first establish unrelativized versions, and then afterward indicatewhat modifications are required.

We begin with □.First we prove □in the following form:Global □Assume V = L. Then there exists ⟨Cµ|µ a singular limit ordinal⟩suchthat:(a) Cµ is CUB in µ(b) ordertype (Cµ) < µ(c) ¯µ ∈Lim Cµ −→C¯µ = C∩¯µ.In the proof we shall take advantage of Jensen’s Σ∗theory, as reformulated inFriedman [94]. For the convenience of the reader we describe that theory here.For simplicity of notation, for limit ordinals µ we let eJµ denote Jα whereωα = µ.

So ORD( eJµ) = µ.Let M denote some Jα, α > 0. (More generally, our theory applies to “accept-able J-models”.) We make the following definitions, inductively.

We order finitesets of ordinals by the maximum difference order: x < y iffα ∈Y where α is thelargest element of (y −x) ∪(x −y).1) A Σ∗1 formula is just a Σ1 formula. A predicate is Σ∗1 (Σ∗1, respectively) ifit is definable by a Σ∗1 formula with (without, respectively) parameters.

ρM1= Σ∗1projectum of M = least ρ s.t. there is a Σ∗1 subset of ωρ not in M and pM1=least p s.t.

A ∩ρM1/∈M for some A Σ∗1 in parameter p (where p is a finite set ofordinals). HM1= HMωρM1 = sets x in M s.t.

M-card (transitive closure (x)) < ωρM1 .For any x ∈M, M1(x) = First reduct of M relative to x = ⟨HM1 , A1(x)⟩whereA1(x) ⊆HM1codes the Σ∗1 theory of M with parameters from HM1∪{x} in thenatural way: A1(x) = {⟨y, n⟩| the nth Σ∗1 formula is true at ⟨y, x⟩, y ∈HM1 }. Agood Σ∗1 function is just a Σ1 function and for any X ⊆M the Σ∗1 hull (X) is justthe Σ1 hull of X.13

(2) For n ≥1, a Σ∗n+1 formula is one of the form ϕ(x) ←→Mn(x) |= ψ, whereψ is Σ1. A predicate is Σ∗n+1 (Σ∗n+1, respectively) if it is defined by a Σ∗n+1 formulawith (without, respectively) parameters.

ρMn+1 = Σ∗n+1 projectum of M = least ρsuch that there is a Σ∗n+1 subset of ωρ not in M and pMn+1 = pMn ∪p where p is leastsuch that A ∩ρMn+1 /∈M for some A Σ∗n+1 in parameter pMn ∪p. HMn+1 = HMωρMn+1 =sets x in M s.t.

M-card (transitive closure (x)) < ωρMn+1. For any x ∈M, Mn+1(x) =(n + 1) st reduct of M relative to x = ⟨HMn+1, An+1(x)⟩where An+1(x) ⊆HMn+1codes the Σ∗n+1 theory of M with parameters from HMn+1 ∪{x} in the natural way:An+1(x) = {⟨y, m⟩| the mth Σ∗n+1 formula is true at ⟨y, x⟩, y ∈HMn+1}.

A good Σ∗n+1function f is a function whose graph is Σ∗n+1 with the additional property that forx ∈Dom(f), f(x) ∈Σ∗n hull (HMn ∪{x}). The Σ∗n+1 hull (X) for X ⊆M is theclosure of X under good Σ∗n+1 functions.Facts.

(a) ϕ, ψΣ∗n formulas −→ϕ ∨ψ, ϕ ∧ψ are Σ∗n formulas(b) ϕΣ∗n or Q∗n (= negation of Σ∗n) −→ϕ is Σ∗n+1(c) Y ⊆Σ∗n hull (X) −→Σ∗n hull (Y ) ⊆Σ∗n hull (X)(d) f good Σ∗n function −→f good Σ∗n+1 function(e) Σ∗n hull (X) ⊆Σ∗n+1 hull (X)(f)There is a Σ∗n relation W(e, x) s.t. if S(x) is Σ∗n then for some e ∈ω,S(x) ←→W(e, x) for all x.

(g) The structure Mn(x) = ⟨HMn , An(x)⟩is amenable. (h) HMn = JAnωρMn where An = An(0).

(i) Suppose H ⊆M is closed under good Σ∗n functions and π : M −→M,M transitive, Range (π) = H and pMn−1 ∈H (if n > 1). Then π preserves Σ∗nformulas: for Σ∗nϕ and x ∈M, M |= ϕ(x) ←→M |= ϕ(π(x)).

And (for n > 1),π(p ¯Mn−1) = pMn−1.Proof of (i)Note that H ∩Mn−1(π(x)) is Σ1-elementary in Mn−1(π(x)). Andπ−1[H ∩Mn−1(π(x))] = ⟨JAωρ, A(x)⟩for some ρ, A, A(x).

But (by induction on n)A = AMn−1 ∩JAωρ, A(x) = An−1(x) ¯M ∩JAωρ. And ρ = ρ ¯Mn−1 using our assumptionabout the parameter pMn−1.

And π−1(pMn−1) = ¯p must be p ¯Mn−1 as ¯M = Σ∗n−1 hull ofH ¯Mn−1 ∪{p ¯Mn−1}.⊣Theorem 4.1. By induction on n > 0 :14

1) If ϕ(x, y) is Σ∗n then ∃y ∈Σ∗n−1 hull (HMn−1 ∪{x})ϕ(x, y) is also Σ∗n.2)If ϕ(x1 · · · xk) is Σ∗m, m ≥n and f1(x), · · · , fk(x) are good Σ∗n functions,then ϕ(f1(x) · · ·fk(x)) is Σ∗m.3) The domain of a good Σ∗n function is Σ∗n4) Good Σ∗n functions are closed under composition.5) (Σ∗n Uniformization) If R(x, y) is Σ∗n then there is a good Σ∗n function f(x)s.t. x ∈Dom(f) ←→∃y ∈Σ∗n−1 hull (HMn−1 ∪{x})R(x, y) ←→R(x, f(x)).6)There is a good Σ∗n function hn(e, x) s.t.

for each x, Σ∗n hull ({x}) ={hn(e, x)|e ∈ω}.Proof. The base case n = 1 is easy (take Σ∗0 hull (X) = M for all X).

Now we proveit for n > 1, assuming the result for smaller n.1) Write ∃y ∈Σ∗n−1 hull (HMn−1∪{x})ϕ(x, y) as ∃¯y ∈HMn−1ϕ(x, hn−1(e, ⟨x, ¯y⟩))using 6) for n −1. Since hn−1 is good Σ∗n−1 we can apply 2) for n −1 to concludethat ϕ(x, hn−1(e, ⟨x, ¯y⟩)) is Σ∗n.

Since the quantifiers ∃e∃¯y ∈HMn−1 range over HMn−1they preserve Σ∗n-ness.2)ϕ(f1(x) · · ·fk(x)) ←→∃x1 · · ·xk ∈Σ∗n−1 hull (HMn−1 ∪{x}) [xi = fi(x)for 1 ≤i ≤k ∧ϕ(x1 · · · xk)]. If m = n then this is Σ∗n by 1).

If m > n thenreason as follows: the result for m = n implies that An(⟨f1(x) · · ·fk(x)⟩) is ∆1 overMn+1(x). Thus Am−1(⟨f1(x) · · ·fk(x)⟩) is ∆1 over Mm−1(x).

So as ϕ is Σ∗m we getthat ϕ(f1(x) · · ·fk(x)) is also Σ1 over Mm−1(x), hence Σ∗m.3) If f(x) is good Σ∗n then dom(f) = {x|∃y ∈Σ∗n−1 hull of HMn−1 ∪{x}(y =f(x))} is Σ∗n by 1).4) If f, g are good Σ∗n then the graph of f ◦g is Σ∗n by 2). And f ◦g(x) ∈Σ∗n−1hull(HMn−1 ∪{x}) since the latter hull contains g(x), f is good Σ∗n and Fact c) holds.5) Using 6) for n −1, let R(x, ¯y) ←→R(x, hn−1(¯y)) ∧¯y ∈HMn−1.

Then R is Σ∗nby 2) for n −1 and using Σ1 uniformization on (n −1) s.t. reducts we can define agood Σ∗n function ¯f s.t.

R(x, ¯f(x)) ←→∃¯y ∈HMn−1R(x, ¯y). Let f(x) = hn−1( ¯f(x)).Then f is good Σ∗n by 4).6) Let W be universal Σ∗n as in Fact f).

By 5) there is a good Σ∗n g(e, x) s.t.∃y ∈Σ∗n−1 hull(HMn−1 ∪{x}) W(e, ⟨x, y⟩) ←→W(e, ⟨x, g(e, x)⟩) (and g(e, x) defined−→W(e, ⟨x, g(e, x)⟩)). Let hn(e, x) = g(e, x).

If y ∈Σ∗n hull ({x}) then for somee, W(e, ⟨x, y′⟩) ←→y′ = y so y = hn(e, x). Clearly hn(e, x) ∈Σ∗n hull ({x}) since15

hn is good Σ∗n.⊣Now we are ready to prove Global □. Assume V = L and let µ be a singularlimit ordinal.

Our goal is to define Cµ, a CUB subset of µ. Let β(µ) ≥µ be the leastlimit ordinal β such that µ is not regular with respect to eJβ-definable functions, andlet n(µ) be least so that there is a good Σ∗n(µ)( eJβ(µ)) partial function from an ordinalless than µ cofinally into µ.

Note that ρβ(µ)n(µ) ≤µ as otherwise such a partial functionwould belong to eJβ(µ), contradicting the leastness of β(µ). Also µ ≤ρβ(µ)n(µ)−1, elsewe have contradicted the leastness of n(µ).For X ⊆eJβ(µ) let H(X) = Σ∗n(µ) hull of X in eJβ(µ).

For some least parameterq(µ) ∈eJβ(µ), H(µ∪{q(µ)}) = eJβ(µ). (“Least” refers to the canonical well-ordering ofL.) Also let α(µ) = S{α < µ|α = H(α∪{q(µ)})∩µ}.

Then (unless α(µ) = S ∅= 0)α(µ) = H(α(µ) ∪{q(µ)}) ∩µ and α(µ) < µ. To see the latter note that for largeenough α < µ, H(α ∪{q(µ)}) contains both the domain and defining parameter fora good Σ∗n(µ) partial function from an ordinal less than µ cofinally into µ.If µ < β(µ) let p(µ) = ⟨q(µ), µ, α(µ)⟩and if µ = β(µ) let p(µ) = α(µ).We are ready to define Cµ.

Let C0µ = {¯µ < µ| For some α, ¯µ = S(H(α ∪{p(µ)}) ∩µ)}. Then C0µ is a closed subset of µ.

If C0µ is unbounded in µ thenset Cµ = C0µ. If C0µ is bounded but nonempty then let µ0 = S C0µ and defineC1µ = {¯µ < µ| For some α, ¯µ = S(H(α ∪{p(µ), µ0}) ∩µ)}.

If C1µ is unboundedthen set Cµ = C1µ. If C1µ is bounded but nonempty then let µ1 = S C1µ and defineC2µ = {¯µ < µ| For some α, ¯µ = S(H(α ∪{p(µ), µ0, µ1}) ∩µ)}.

Continue in thisway, defining Ckµ for k ∈ω until Ckµ is unbounded or empty. Note that α0 > α1 >· · · where αk is greatest so that S(H(αk ∪{p(µ), µ0, · · · , µk−1}) ∩µ) = µk, sinceαk ∈H(αk ∪{p(µ), µ0, · · · , µk−1, µk}).

So for some least k(µ) ∈ω, Ck(µ)µis indeedunbounded or empty. If Ck(µ)µis unbounded then set Cµ = Ck(µ)µ.If Ck(µ)µ= ∅then we choose Cµ to be an ω-sequence cofinal in µ, codingapproximations to the structure eJβ(µ), as follows.

(This is necessary to establishRelativized □(c).) Note that H = H({p(µ), µ0, · · · , µk(µ)−1}) is cofinal in µ sinceCk(µ)µ= ∅.

Assume first that n(µ) = 1, when Cµ is more easily described. ThenH is also cofinal in β(µ), else H ∈eJβ(µ) and µ is singular inside eJβ(µ).

Let h =h1(e, x) be the canonical good Σ∗1 Skolem function for eJβ(µ), so H = {h(e, p)|e ∈ω}where p = {p(µ), µ0, · · · , µk(µ)−1}. Let ¯σn = max({h(e, p)|e < n} ∩µ) and σn =16

max({h(e, p)|e < n} ∩β(µ)). Then Cµ = {δ0, δ1, · · ·} where δn is an ordinal codingTC(Σ∗1 hull (¯σn ∪{p}) restricted to σn), where TC denotes “transitive collapse”.By the Σ∗1 hull of X restricted to σn we mean the closure of X under hδn, obtainedby interpreting the Σ∗1 definition of h in eJσn.Now suppose n(µ) > 1, Ck(µ)µ= ∅.

Then if ρ(µ) denotes ρµn(µ)−1, H is cofinalin ρ(µ), else H ∈eJρ(µ) and µ is singular in eJρ(µ). Let h be the canonical goodΣ∗n(µ) Skolem function for eJρ(µ) and let p = {p(µ), µ0, · · · , µk(µ)−1}.

Let ¯σn =max({h(e, p)|e < n}∩µ), σn = max({h(e, p)|e < n}∩ρ(µ)). Then Cµ = {δ0, δ1, · · ·}where δn is an ordinal coding TC(Σ∗n(µ) hull (¯σn ∪{p}) restricted to σn).

The Σ∗n(µ)hull of X restricted to σn is the closure of X under hσn, obtained by replacing the(n(µ) −1) st reduct Mn(µ)−1(x) by Mn(µ)−1(x) ↾σn in the Σ∗n(µ) definition of h.(Recall that Mn(x) = ⟨JAnωρMn , An(x)⟩; by Mn(x) ↾σ we mean ⟨eJAnσ , An(x) ∩eJAnσ ⟩. )Clearly Cµ is CUB in µ, and by the same argument used to justify α(µ) < µ,the ordertype of Cµ is less than µ.

(These facts are obvious when Ck(µ)µ= ∅.) So toprove Global □we only need to check coherence: ¯µ ∈Lim Cµ −→C¯µ = Cµ ∩¯µ.Lemma 4.2.

¯µ ∈Ckµ −→Ck¯µ = Ckµ ∩¯µ.Proof. First suppose that k = 0.

Given ¯µ ∈C0µ we can choose α < ¯µ such that¯µ = S(H(α∪{p(µ)})∩µ), where H is the operation of taking the Σ∗n(µ) hull. Also letρ = S(H(α∪{p(µ)})∩ρµn(µ)−1).

Let π : eJ ¯β −→eJβ(µ) be the inverse to the transitivecollapse of H = Σ∗n(µ) hull (¯µ∪{p(µ)}) restricted to ρ. Note that any x ∈H belongsto H(µ′, ρ′) = Σ∗n(µ) hull (µ′ ∪{p(µ)}) restricted to ρ′, for some µ′ < ¯µ, ρ′ < ρ andµ′, ρ′ can be chosen to be in H. It follows that H ∩µ = ¯µ and therefore whenµ < β(µ), π(¯µ) = µ.

Also note that Σ∗n(µ)−1 hull (ρ ∪{p(µ)}) ∩ρµn(µ)−1 = ρ,so H is closed under good Σ∗n−1 functions. It follows that π :eJ ¯β −→eJβ(µ) isΣ∗n−1-elementary and ¯µ is Σ∗n(µ)−1( eJ ¯β)-regular, Σ∗n(µ)( eJ ¯β)-singular.

So ¯β = β(¯µ),n(µ) = n(¯µ). Also π(q(¯µ)) = q(µ).

Since α(¯µ) < α it must be that α(¯µ) = α(µ). Soπ(p(¯η)) = p(µ).

Now it is easy to see that C0¯µ = C0µ ∩¯µ.Now suppose k = 1. The above argument shows that ¯µ ∈C1µ −→C0¯µ = C0µ ∩¯µand hence, since µ0 < ¯µ, ¯µ0 = µ0.

Now again, the above argument shows thatC1¯µ = C1µ ∩¯µ. The general case k ≥0 now follows similarly.⊣Coherence now follows easily: if ¯µ ∈Lim Cµ and Cµ = Ckµ then by Lemma 4.2,Ck¯µ = Ckµ ∩¯µ is unbounded in ¯µ so C¯µ = Ck¯µ and we’re done.

If Ckµ = ∅for some k17

then lim Cµ = ∅so coherence is vacuous.To establish the appropriate relativized form of □we need:Lemma 4.3. Suppose π :⟨eJ¯µ, ¯C⟩Σ1−→⟨eJµ, Cµ⟩.

Then C = C¯µ and π extendsuniquely to a Σ∗n(µ)-elementary ˜π :eJβ(¯µ) −→eJβ(µ) such that p(µ) ∈Rng ˜π.Proof. First suppose that Cµ = Ckµ for some k. For µ′ ∈Cµ form H(µ′) as Hwas formed in the proof of Lemma 2 for ¯µ.

Then π(µ′) :eJβ(µ′) −→eJβ(µ) withrange H(µ′) is Σ∗n(µ)−1-elementary and eJβ(µ) = S{H(µ′)|µ′ ∈Cµ}. And π(µ′) ↾µ′ = id ↾µ′, π(µ′)(p(µ′)) = p(µ).

Now let X = Range(π) and form eX = Σ∗n(µ)hull (X ∪{p(µ)}) in eJβ(µ). If y ∈eX then for some µ′ ∈Cµ, y = π(µ′)(y′) wherey′ ∈Σ∗n(µ) hull ((X ∩eJµ′) ∪{p(µ′)}).

In particular if y ∈eJµ then y ∈Σ∗1 hull (X)in ⟨eJµ, Cµ⟩= X. So the inverse to the transitive collapse of eX = ˜π is a Σ∗n(µ)-elementary embedding extending π, with p(µ) in its range.

If ˜π :eJ ¯β −→eJβ(µ)then ¯µ = ˜π−1(µ) is singular via a Σ∗n(µ)( eJ ¯β) partial function since either Σ∗n(µ) hullof µ′ ∪{p(µ)} in eJβ(µ) is unbounded in µ for some µ′ < S(Rng(π) ∩µ), in whichcase we can assume µ′ ∈Rng π and by Σ∗n(µ)-elementary of ˜π we’re done, or if notµ∗= µ∩Σ∗n(µ) hull (µ∗∪{p(µ)}) in eJβ(µ) where µ∗= S(Rng π∩µ), contradicting thedefinition of α(µ). Since ¯µ is Σ∗n(µ)−1( eJ ¯β)-regular, we get ¯β = β(¯µ), n(µ) = n(¯µ).Then the Σ∗n(µ)-elementarity of ˜π guarantees that C = C¯µ.

The uniqueness of ˜πfollows from the fact that eJβ(¯µ) = Σ∗n(µ) hull (¯µ ∪{p(¯µ)}) and ˜π ↾¯µ is determinedby π.If Ckµ = ∅for some k then Cµ was defined as a special ω-sequence cofinal in µ.That definition was made precisely to enable the preceding argument to also applyin this case.⊣Relativized □Let S = S{Sα|α an infinite cardinal}. There exists ⟨Cs|s ∈S⟩such that Cs ∈As and:(a) Cs is closed, unbounded in µ0s, ordertype (Cs) ≤α(s).If |s| is a successor ordinal then ordertype (Cs) = ω.

(b) ν ∈Lim(Cs) −→for some η < |s|, ν = µ0s↾η and Cs ∩ν = Cs↾η. (c)Let π :⟨A, C⟩Σ1−→⟨A0s, Cs⟩and write crit(π) = ¯α, A = L¯µ[A, ¯s∗].

Ifπ(¯α) = α(s) then L[A, ¯s∗] ⊨|¯s| is not a cardinal > ¯α and(c1) C ∈Lµ[A, ¯s∗] where µ is the least p.r. closed ordinal greater than ¯µ s.t.18

Lµ[A, ¯s∗] ⊨card(|¯s|) ≤¯α. (c2) π extends to π′ : A′Σ1−→A′s where A′ = Lµ′[A, ¯s∗], µ′ = largest ordinaleither equal to ¯µ or s.t.

ω · µ′ = µ′ and Lµ′[A, ¯s∗] ⊨|¯s| is a cardinal greater than ¯α. (c3) If ¯α is a cardinal and π(¯α) = α then A = A0¯s and C = C¯s.Relativized ⋄Let E = all s ∈S such that ordertype (Cs) = ω.

There exists⟨Ds|s ∈E⟩such that Ds ⊆A0s and:(a)D ∈bAs ̸= A0s, D ⊆A0s −→{ξ < |s|s ↾ξ ∈E, Ds↾ξ = D ∩A0s↾ξ} isstationary in bAs. (b) Ds is uniformly Σ1-definable as an element of A′s.

(c) If A′s ⊨α++ exists then Ds = ∅.We now make the necessary modifications to obtain Relativized □. First, if µis a singular limit ordinal and eJµ |= α is the largest cardinal then we thin out Cµto give it ordertype ≤α : By induction on limit ¯µ ≤µ define C∗¯µ as follows.

For¯µ ≤α, C∗¯µ = ¯µ. Otherwise C∗¯µ = {ith element of C¯µ|i ∈C∗¯µ0 where ¯µ0 = ordertype(C¯µ)}.

This defines C∗µ. It is easily verified that the C∗µ enjoy all the properties ofthe Cµ except they are only defined when µ is a singular limit ordinal such thateJµ |= There is a largest cardinal.

In addition, ordertype C∗µ ≤α(µ), the largestcardinal of eJµ.Now suppose V = L, α is a cardinal, s ∈Sα, |s| > α and s is a 0-string, meaningthat s(µ) = 0 for all η ∈Dom(s). Then we can choose our predicate A = ∅, defineCs = C∗µ0s and Relativized □will hold for such 0-strings.

The final comment is thatall we have done will relativize to arbitrary strings s ∈Sα, defined relative to anarbitrary predicate A ⊆ORD, Hα = Lα[A] for all cardinals αNow we turn to Relativized ⋄. Again we begin with a nonrelativized version.Let α be a cardinal and assume V = L.⋄on α+ Let E = all µ < α+ s.t.

Cµ has ordertype ω. There exists ⟨Dµ|µ ∈E⟩s.t.Dµ ⊆eJµ and:(a) If D ⊆eJα+ then {µ ∈E|D ∩eJµ = Dµ} is stationary in α+.

(b) Dµ is uniformly Σ1 definable as an element of eJβ′(µ) where β′(µ) = largestβ s.t. either β = µ or ωβ = β and eJβ |= µ is a cardinal greater than α.

(c) If eJβ′(µ) |= α++ exists then Dµ = ∅.19

Proof. For µ ∈E let Dµ = ∅if eJβ′(µ) |= α++ exists and otherwise let ⟨Dµ, Fµ⟩beleast in eJβ′(µ) such that Fµ is CUB in µ and ¯µ ∈Fµ −→¯µ /∈E or D¯µ ̸= Dµ ∩eJ¯µ.If ⟨Dµ, Fµ⟩doesn’t exist let Dµ = ∅.

Properties (b), (c) are clear. To prove (a),suppose it fails and let ⟨D, F⟩be least in eJα++ such that D ⊆eJα+, F is CUB inα+ and µ ∈F −→µ /∈E or Dµ ̸= D ∩eJµ.

Let σ be least such that ωσ = σand ⟨D, F⟩∈eJσ. Then eJσ |= α+ is the largest cardinal.

Let H = Σ1 Skolem hullof {α+} in eJσ and µ = S(H ∩α+). Then eJβ′(µ) is the transitive collapse of theΣ1 Skolem hull of µ ∪{α+} in eJσ; let π :eJβ′(µ) −→eJσ have range equal to thelatter hull.

Then we have a contradiction provided µ ∈E. But the fact that H isunbounded in µ implies that C0µ = ∅so ordertype (Cµ) is ω and µ ∈E.⊣Now as with Relativized □, if V = L and α is a cardinal, s ∈Sα is a 0-string inE, |s| > α then we can choose A = ∅and define Ds = Dµ0s where the latter comesform ⋄on α+.

Finally, relativize everything to arbitrary reshaped strings s ∈Sαand an arbitrary predicate A ⊆ORD, Lα[A] = Hα for all cardinals α.This completes the proof of Relativized □and ⋄.ReferencesBeller-Jensen-Welch [82] Coding the Universe, Cambridge University Press.Friedman [87] Strong Coding, Annals of Pure and Applied Logic.Friedman [94] Jensen’s Σ∗Theory and the Combinatorial Content of V = L, toappear, Journal of Symbolic Logic.Jensen [72] The Fine Structure of the Constructible Hierarchy, Annals of Mathe-matical Logic.20


출처: arXiv:9211.204원문 보기

Subscribe to koineu.com

Don’t miss out on the latest issues. Sign up now to get access to the library of members-only issues.
jamie@example.com
Subscribe